Human peripheral lymphocytes were infected with the human-avirulent Microtus 201 strain or the fully virulent Antiqua strain 141 each at a multiplicity of infection (MOI) of 10. The infected lymphocytes harvested at 2, 4, and 8 h post-infection (hpi) were subjected to the total RNA isolation and RNA sequencing (RNA-seq) analysis. Clean reads from the RNA-seq analysis were mapped to the annotated human genome (NCBI36/hg18). Each library comprised around 4.5 × 10⁷ clean reads, 62.83% of which were on average mapped uniquely to the human reference genome (Table 1). Gene expression was calculated using the fragments per kilobase of gene per million fragments mapped (FPKM) method (Mortazavi et al., 2008), and the genes whose expression levels were altered by over 2-fold in expression level with false discovery rates (FDRs) ≤ 0.001 were defined as being the differentially expressed genes (Benjamini and Yekutieli, 2005).

First, the transcriptomes of the peripheral lymphocytes infected with strains 201 or 141 were compared with that of the uninfected lymphocytes, respectively. Differentially expressed genes were identified according to the aforementioned criteria and the numbers of genes at each time point post-infection were calculated. More than 260 genes were differentially expressed at 2 hpi between the 201-infected and uninfected lymphocytes, and the number increased greatly as the infection period proceeded and exceeded 1,600 at 8 hpi (Figure 1A). In sharp contrast, the transcriptome of the 141-infected peripheral lymphocytes differed very little from that of the uninfected cells at 2 hpi, and fewer than 20 genes were found to be differentially expressed, although the numbers of differentially expressed genes at 8 hpi in the 141- and 201-infected cells were high and comparable (Figure 1B).

Further comparative analysis showed once again that the most remarkable gene expression changes between the 201- and 141-infected peripheral lymphocytes occurred at the earliest time point that was analyzed (Figure 1C). The differentially expressed genes were relatively fewer at 8 hpi although they were quite abundant in the 201-infected (Figure 1A) and the 141-infected (Figure 1B) lymphocytes in comparison with the uninfected lymphocytes control. These data demonstrate that the initial contact by the human peripheral lymphocytes with the two Y. pestis strains triggered two obviously distinct responses, one of which could possibly facilitate efficient recognition of 201 bacilli and boost strong immune responses, while the other failed to do so in the case of infection with Y. pestis strain 141. It appeared that the 141-infected lymphocytes were strongly inhibited in their ability to deploy an inflammatory immune response because of a vital virulence mechanism present in strain 141, but absent in strain 201, which might originate from some as yet uncharacterized difference(s) between the two strains.

To validate the RNA-seq data, primer pairs for 23 genes were designed for quantitative RT-PCR analysis (Supplementary Table 1), among which 12 genes were up-regulated, 7 genes down-regulated and 3 unchanged in the 201-infected lymphocytes in comparison with the uninfected control according to the RNA-seq results. qRT-PCR was performed using the same RNA samples used in RNA-seq library construction as templates, 16s RNA was the reference gene for data normalization, and the log2 ratios of the target gene concentrations in the 201-infected lymphocytes to those of the uninfected control were calculated. The correlation efficiencies (R²) between the qRT-PCR and RNA-seq analysis results were over 0.90 for the RNA samples collected at 2 and 8 hpi, and 0.83 for the samples collected at 4 hpi (Supplementary Figure 1 and Table 2). These results indicate that the RNA-seq data described in this study are highly reliable.

To characterize the host transcriptomic response to the human avirulent Microtus strain 201, genes differentially expressed in 201- or 141-infected human peripheral lymphocytes, compared with the uninfected cells, were subjected to the pathway enrichment analysis using the DAVID 6.7 bioinformatics tool (http://david.abcc.ncifcrf.gov) (Huang da et al., 2009). Consistent with the observation that the genes differentially expressed between 201- and 141-infected lymphocytes peaked at 2 hpi, several pathways critical for inducing the intense immune responses required for invading pathogen clearance were significantly enriched in the 201- but not the 141-infected lymphocytes (p < 0.05, Fisher's exact test followed by the Bonferroni multiple testing correction) at this time point. These pathways included the cytokine-cytokine receptor interaction pathway, the chemokine signaling pathway, the TLR signaling pathway, and the Jak-STAT signaling pathway, among others (Table 2). Figure 2 shows the heat maps for some of the differentially enriched pathway in details. Strikingly, at the later infection time points (4 and 8 hpi), the pathway enrichment analysis revealed a significant difference between the 201- and 141-infected lymphocytes in the lysosome pathway, the Fcγ receptor-mediated phagocytosis pathway and the extracellular-matrix (ECM) receptor pathway, suggesting that the phagocytosis and lysosome maturation process could be differentially modulated in the lymphocytes infected with the two different strains.

Pathogens invading the mammalian host are recognized by pathogen recognition receptors (PRRs) via sensing of the conserved ligands on microorganisms (Takeuchi and Akira, 2010), and TLRs represent the major PRRs responsible for bacterial pathogen recognition (Krishnan et al., 2007). The signaling pathways triggered by TLRs exhibit distinct characteristics in several key molecules (Supplementary Figure 2A). CD14 protein binds to lipopolysaccharide (LPS), which after transfer to the TLR4/MD-2 complex triggers the downstream signaling required for the production of pro-inflammatory cytokines and type I interferon (Leeuwenberg et al., 1994; Ishihara et al., 2004). TLR adaptor molecule 2 (TICAM2) is an adaptor that transduces signals upon ligand binding to TLR4. Interferon regulatory factor 5 (IRF5) is a transcriptional factor involved in activation of interferon and the immune system. TLR8 recognizes single-stranded RNA from both viruses and bacteria (Sarvestani et al., 2012). The expression levels of CD14, TLR8, TICAM2, and IRF5 were much higher in the 201-infected lymphocytes than in their 141-infected counterparts at 2 hpi, a finding further confirmed by the qRT-PCR analysis (Supplementary Figure 2B), although their expression levels after this time point gradually became comparable in the 201- and 141-infected lymphocytes.

As a facultative intracellular pathogen, Y. pestis is readily taken up by phagocytes but the pathogen becomes phagocytosis-resistant shortly after its initial intracellular survival and replication (Pujol et al., 2009; Connor et al., 2015). Thus, in systemic infection of the mammalian hosts, the survival of Y. pestis in host macrophages at the early stage of infection are critical for the later disease progress. Our results revealed that the differentially expressed genes were significantly enriched in the lysosome pathway in 201-infected lymphocytes, and 30 genes were significantly changed at 8 hpi (Table 2) but not in the 141-infected cells. Although we saw no significant enrichment in this pathway at 2 hpi, several key genes (e.g., ATP6V0A1, CYSZ, CTSD and SLC11A1) involved in lysosome formation and maturation were expressed at higher levels in the 201-infected lymphocytes at 2 hpi than in the 141-infected cells. ATP6V0A1 encodes a vacuolar ATPase that mediates acidification of intracellular organelles, and vacuolar acidification is necessary for zymogen activation in the lysosome (Saw et al., 2011). CTSD protein exhibits pepsin-like activity, whereas CTSZ is a lysosome cysteine protease that has been shown to have lysosomal degradative capacity (del Cerro-Vadillo et al., 2006; Liu et al., 2014; Tranchemontagne et al., 2016). SLC11A1 functions as a divalent transition metal transporter involved in host resistance to infection with Mycobacterium tuberculosis and M. leprae, and also is associated with some inflammatory diseases (Govoni et al., 1999). qRT-PCR analysis of the aforementioned genes expressed in the 201- and 141-infected lymphocytes (Supplementary Figure 3) confirmed the RNA-seq results that the genes were expressed at higher levels in the 201-infected lymphocytes than in their 141-infected counterparts, suggesting that the lysosome pathway response differed between the infection with strains 201 and 141.

Next, we sought to compare cytokines and chemokines secretion in lymphocytes infected with the two different strains. Culture supernatants from the infected lymphocytes were collected and analyzed for IL-8, IL-1β, IL-6, IL-10, TNF-α, IL-12p, Interferon-γ (IFN-γ), IL-17A using BD^TM CBA Kits. Only the cytokines and chemokines that showed statistically significant differences (p < 0.05) were taken into account in the following analysis (Figure 3). The immunological measurements results showed that the 201-infected lymphocytes secreted much more IFN-γ and IL-1β than those infected with strain 141. IFN-γ is critical for protection against plague and treatment with IFN-γ greatly enhances the bactericidal activity of macrophages against the Y. pestis bacterium (Pujol et al., 2005; Parent et al., 2006). Y. pestis is known to be able to suppress the production of inflammatory cytokines IL-1β to promote its infection process (Ratner et al., 2016). Production of IFN-γ and IL-1β, both of which paly essential roles in host immune responses, appear to be significantly inhibited in the presence of strain 141 in comparison with the human avirulent strain 201. By contrast, secretion of the immunosuppressive cytokine Interleukin-10 (IL-10) was significantly higher in the 141-infected lymphocytes, unlike their 201-infected counterparts, which accords with previous reports that IL-10-deficient mice are resistant to Y. pestis (Turner et al., 2009).

The immediate up-regulation of the proteins involved in lysosome activation in 201-infected human lymphocytes promoted us to determine their expression levels after infection with Y. pestis strain 201. Human macrophage-like U937 cells primed with phorbol myristate acetate (PMA) and murine macrophage RAW264.7 cells were infected with Y. pestis strain 201, and the infected cells collected at 2, 4, 8 hpi were lysed for the immunoblotting detection of ATP6V0A1, SLC11A1, CTSD and CTSZ. Primed U937 cells possess major macrophage characteristics and can therefore resemble lysosome activation quite well in the infected host phagocytes (Koren et al., 1979; Grabenstein et al., 2006). With the exception of ATP6V0A1, whose lack of detection was probably related to its low abundance, all of the other three proteins were successfully detected (Figure 4). The levels of CTSD seemed to be slightly increased in the U937 cells after infection with the 201 strain, and the levels of CTSZ showed no significant alterations during the infection. Because experiments involving human virulent Y. pestis strains must be performed at biosafety level 3 laboratory, we were unable to obtain the expression levels of these proteins in 141-infected lymphocytes. However, interestingly, when the RAW264.7 cells were infected with Y. pestis strain 201, both the basal expression of CTSD and those after infection were quite low compared with that of the U937 cells, indicating that very low expression of this protease occurred under these conditions. The expression levels of SLC11A1 in both U937 and RAW264.7 cells were significantly increased after infection, and the induction of SLC11A1 in RAW264.7 seemed to be even stronger. The different expression patterns of CTSD and SLC11A1 in U937 and RAW264.7 hinted us that the lysosome activity toward Y. pestis strain 201 could be distinctive, which is likely relevant to the host-specific pathogenicity of Microtus strains of Y. pestis. However, no significant alteration in CTSZ expression was seen in both U937 and RAW264.7 cells although its' transcript level showed > 10-fold abundance at 8 hpi in the human lymphocytes, probably related to some post-transcriptional regulatory mechanisms.

Because U937 and RAW264.7 cells seemed to respond to 201 and 141 strains differentially in respect of lysosome activation, we wanted to further characterize the interactions of Y. pestis strain 201 with these cell lines. U937 and RAW264.7 cells were infected with the bacteria and their survival percentages were measured using a gentamycin protection assay followed by plating cell lysates on agar plates. The U937 cells were stimulated with PMA (100 ng/mL) for 48 h before they were infected. Both U937 and RAW264.7 cells were infected with the Y. pestis 201 strain at a MOI of 1 or 5, and the infections were allowed to proceed for 32 h. The number of live bacteria in U937 and RAW264.7 cells at 0.5 h hpi were designated as the initial value and the survival percent at 2, 4, 8, 20, and 32 hpi were calculated by dividing the corresponding live bacterial cell number to that at 0.5 hpi. The percentage survival of Y. pestis bacilli decreased drastically to about 25–30% in both types of cells at 2 hpi at a MOI of 1. This percentage was sustained in RAW264.7 cells at a relatively steady level, dropping only 4% from 2 to 8 hpi; however, the bacteria inside the U937 cells were cleared much faster, as shown by the prominent decrease from 32 to 9% during this time period. At 32 hpi, no living 201 bacteria were detected in the U937 cells, whereas 10.6% of the bacteria were still alive in the RAW264.7 cells (Figure 5). When the cells were infected with Y. pestis 201 at an MOI of 5, similar results were found, where the percentage survival of the 201 bacteria in the RAW264.7 cells was significantly higher than that in the U937 cells, indicating that human macrophages can clear the Y. pestis Microtus strain more efficiently than the murine macrophages.

Peripheral lymphocytes were isolated from the whole blood collected from 8 healthy donors using Ficoll-Paque Plus (GE Healthcare, USA). The purified lymphocytes were maintained in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS) and 2 mmol/L of L-glutamine at 37°C in a 5% CO₂ incubator. Y. pestis 141 strain is a biovar Antiqua strain, highly virulent in mice and human. Y. pestis strain 201 belongs to the biovar Microtus and is highly virulent in mice but avirulent in humans. The glycerol-preserved strains were inoculated into 5 mL of brain heart infusion (BHI) medium and allowed to grow at 26°C for 24 h. The strains recovered were subjected to three consecutive passages and the third passage cultures were grown until they approached an optical density the (OD) _600nm value equal to 1. Bacterial cultures were centrifuged and resuspened in RPMI-1640. The purified human peripheral lymphocytes were infected with the bacterial suspensions of 141 or 201 strains at an MOI of 10, and the infected cells were centrifuged briefly to promote the adhesion of the bacteria to the lymphocytes before incubation at 37°C in a 5% CO₂ incubator. Gentamycin at 100 μg/ml was added after 0.5 h of infection to kill the extracellular bacteria in order to inhibit the proliferation of the bacteria during infection. At the 0.5, 2, 4, and 8 hpi, the culture supernatants from the 201- or 141-infected lymphocytes were collected and centrifuged at 4.000 rpm for 5 min to remove the floating bacteria and cell debris, and the cleared supernatants were stored in small aliquots for the cytokine measurements. The infected lymphocytes were collected and subjected to total RNA isolation.

Total RNAs were purified from the lymphocytes infected with strains 201 or 141 using a TRIzol reagent (Invitrogen, Carlsbad, CA) and the isolated RNAs were treated with RNAse free DNase I (Thermo Fisher Scientific, Waltham, MA). The RNA integrity number, rRNA ratio (28S/18S), and the total RNA concentration in the samples were determined according to the previous descriptions (Du et al., 2014; Yang et al., 2017). Sequencing libraries were prepared using the Illumina Truseq RNA sample preparation kit according to the manufacturer's protocol. Briefly, the first-strand cDNA was generated by First Strand Master Mix and SuperScript II reverse transcription using random primers and the second-strand cDNA was synthesized using DNA polymerase I and RNase H. The fragmented cDNA molecules were end-repaired, purified, and then ligated with PloyA. The sequencing was performed using Illumina HiSeq 2000 (Illumina Inc., USA). The original image data is converted into “raw reads” via base calling. To obtain clean reads of acceptable quality, raw data filtering and quality control were performed to remove the adaptors-containing reads, unknown bases amounting to more than 10%, or low quality bases. The Burrows-Wheeler alignment tool was used to map the clean reads to the human reference genome (NCBI36/hg18) (Li and Durbin, 2009), and Bowtie was used to map to the reference genes (Langmead et al., 2009). Genes and isoform expression levels were quantified by the RSEM software package. We used the FPKM method to calculate the gene expression levels (Mortazavi et al., 2008). The FDR was computed by dividing the number of falsely discovered genes at a given p-value by the number of statistically significant differentially expressed genes comparing the sample to the control at the same p-value (Benjamini and Yekutieli, 2005). The criteria used for differentially expressed genes (DEG) were FDR ≤ 0.001 and log2 (FPKM in the test sample/FPKM in the reference) ≥ 1. All the data discussed here have been deposited to GEO at NCBI with the accession number of GSE121084.

To confirm gene expression levels by qRT-PCR, RNA samples that were subjected to the RNA-sequencing were used as templets and the qRT-PCR analysis were performed based on SYBR Green I fluorescence using Roche Light Cycler 480. RNA samples from the uninfected, 201- or 141-infected lymphocytes collected at the 2, 4 and 8 hpi were used as templates with primer pairs to amplify the corresponding target genes (Supplementary Table 2). Correlations between the sequencing data and the qRT PCR results were calculated using linear regression methodology.

The cytokines present in the culture supernatants from human lymphocytes infected with 201 or 141 strains for 2, 4, and 8 h were detected using the BD™ Cytometric Bead Array (CBA) Human Inflammatory Cytokine Kit and Human Th1/Th2/Th17 Kit (BD Biosciences, San Jose, CA). The cytokines concentrations were measured by a Becton-Dickinson FACS Caliber flow cytometer. Cytokine levels were processed using Graphpad Prism 5.0, and the statistical significance of the differences in the cytokine levels between the different groups of samples were analyzed by two-way ANOVA analysis followed by Bonferroni post-tests.

U937 and RAW264.7 cells were maintained in RPMI 1640 medium containing 10% FBS and 2 mmol/L of L-glutamine at 37°C in a 5% CO₂ incubator. U937 cells were primed with PMA (100 ng/mL) for 48 hours before infection. U937 and RAW264.7 cells were seeded onto 24-well plates at a concentration of 4 × 10⁵/mL the day before infection. Y. pestis strains 201 and 141 were grown in BHI until they reached OD_600nm = 1, and each were collected separately by centrifugation and resuspendend in RPMI 1640. The U937 cells or RAW264.7 cells were then infected with strains 201 or 141 at an MOI of 1 or 5, and gentamycin at 100 μg/ml was added to the cells to kill the extracellular bacteria after 0.5 h of infection. At 0.5, 2, 4, 8, 12, 20, 32 hpi, the culture medium from each well was decanted, the cells were thoroughly washed in phosphate-saline buffered saline (PBS), and the infected cells were lysed by addition of sterile H₂O containing 0.1% Triton X-100 for 15 min at room temperature to release the engulfed bacteria. The living bacteria were counted by plating the diluted cell lysate onto to the agar plates in triplicate. This experiment was performed independently three times as biological triplicates and a representative result is shown as the mean ± SD (n = 3).

U937 cells were PMA-primed before they were infected with the Y. pestis strain as described above. U937 cells and RAW264.7 cells were maintained, seeded and infected with Y. pestis strains 201 and 141 at an MOI = 10, according to the method described in the survival capability assays. Gentamycin was added to the cultures to inhibit the over proliferation of the bacteria at 0.5 hpi. U937 and RAW264.7 cells were harvested at 2, 4, and 8 hpi, washed with ice-cold PBS, and lysed with lysis buffer (50 mM Tris-HCl [pH 7.4], 150 mM NaCl, 1% TritonX-100, 1% sodium deoxycholate, 2 mM sodium pyrophosphate, 1 mM EDTA) supplemented with a protease inhibitor complete cocktail (Roche, Basel, Switzerland). Sodium dodecyl sulfate (SDS) loading buffer was added into the cell lysate, which was boiled and then separated by 12% SDS-polyacrylamide gel electrophoresis (SDS-PAGE), followed by transfer onto a polyvinylidene fluoride membrane (GE Healthcare, Piscataway, NJ, USA). Specific proteins were detected using antibodies against Actin, CTSZ, CTSD and SLC11A1(Santa Cruz Biotechnology, Dallas, TX).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.